AI Art Weekly #9
Welcome to issue #9 of AI Art Weekly.
Today our newsletter launched on Product Hunt. Would mean the world to me if you could leave a comment and give me some love on the page as it helps the newsletter grow and get more exposure 🙏
That being said, here are some of this weeks highlights:
- Thought-To-Image paper, new Text-To-3D models and Deforum v0.6 released.
- Interview with Black Label Art Cult.
- DreamArtist – an automatic1111 img2img extension similar to MJv4 and Nitro-Diffusion – the first multi style model trained from scratch got released.
Let’s jump in.
Cover Challenge 🎨
This weeks “ comic ” challenge got 61 submissions from 38 artists and the community decided on the final winner. It was the closest vote yet and @GothlyticalArt came out on top as the winner. Congratulations! And as always a big thank you to everyone who contributed!
The theme for the next challenge is “ forestpunk ”. Think mystical, self-sustainable, everything forest. Checkout the aesthetics wiki for more info. Prize is another $50. I made some minor changes to the Rulebook. Images can be submitted here.
I’m looking forward to all of your submissions 🙏
If you want to support the newsletter, this weeks cover is available for collection on objkt as a limited edition of 10 for 2.50ꜩ a piece. Thank you for your support 🙏😘
Reflection: News & Gems
Thought-To-Image 🧠
The most exciting paper announcement this week goes to MinD-Vis which reconstructs highly plausible images from brain recordings. Forget these user unfriendly and cumbersome interfaces that require you to write text or speak words. In the near future what you think about will manifest itself as a picture. Not sure if that will be a good thing per se, but certainly a mind-blowing own 😉.
Mind-Vis examples
New Text-to-3D models
“Latent-NeRF for Shape-Guided Generation of 3D Shapes and Textures” is collection of three models that enable the generation of 3D objects with textures either by providing a pure text prompt, a basic 3D shape or input images. There is no available code yet, but according to the GitHub repo it’s being finalized and coming soon.
Latent-NeRF texture examples based on a given mesh.
InstructPix2Pix
Imagine being able to edit existing images by simply saying “add fireworks to the sky” or “replace mountains with city skyline” instead of constructing a whole paragraph of words. InstructPix2Pix is working on this ability and it looks like we’ll soon be able to use it.
InstructPix2Pix example using natural instructions to edit images
VQ-Diffusion
Did you know that Microsoft has released an open source diffusion model called VQ-Diffusion? Me neither. Generation quality isn’t comparable to Stable Diffusion yet. But apparently it’s based on a new innovating architecture. There is a HuggingFace demo in case you want to try it out.
“A man with a pineapple head”. Left is VQ. Right is SD.
Deforum v0.6
The Deforum animation notebook received a new update, adding new conditioning features, xformers efficient attention (which means faster frame rendering), masking for 2D and 3D video modes and other new things. Oh, and remember DPM Solver++ from two weeks ago? It’s now available within the Deforum notebook and web extension and produces some nice results with only 20 steps.
@DIGIMONSTER1006 created a short two part science fiction comic based on the “comic” cover challenge this week. Part1 and Part2 can be found on Twitter.
@bioinfolucas showed us this week how to easily create animated 2D game assets using Stable Diffusion and sketch.metademolab.com.
@gd3kr developed an AI bot that answers all of grandma’s technology questions so you don’t have to spend an hour on FaceTime trying to get her to click the right buttons. For anyone who still had any doubts: I asked her if AI art is art and she said yes. You can live in peace now.
.jpg)
Imagination: Interview & Inspiration
In this weeks interview of AI Art Weekly we talk to Black Label Art Cult. An artist and serial entrepreneur who not only makes waves with his art, but also by speaking up for our AI art community against false narratives and the hate some of us had to endure over the last few weeks.
[AI Art Weekly] Hey BL, what’s your background and how did you get into AI art?
I’m a serial entrepreneur and the founder of 5 businesses/companies, I started my first business when I was 19 and since then I have ran multiple agencies as well as other startups, I’ve worked in international communication companies, media companies, and have worked with over two hundred clients stretching across the globe and 14 countries. But at my core I’ve always been an artist. Traditionally, I’m an abstract expressionist dealing mostly with oil and acrylic on canvas. I’ve been in the creative field most of my career as well, with a strong expertise in brand identity, graphic design, UI design, web design, and a lot more. I first came across text to image AI generative art about a year ago. I immediately hired a developer and created my own proprietary AI, and used that for the first half of the year of 2022. The AI was never open to the public, but it was an evolution of VQGAN and VDIFF code bases. With the release of MidJourney and Stable Diffusion my process has evolved utilizing those tools to further push my creative process.
An old sample of BL’s own AI model
[AI Art Weekly] Do you have a specific project you’re currently working on? What is it?
This week I’m actually beginning to drop a new series called False Idols, on KnownOrigin. The series deals with surrealist landscapes of post-revolutionary regret. The iconography represents a societal downfall that was brought upon by a revolt against a global ideology that seemed to be oppressive, but the scenes depicted express a sincere sense of regret through the showing of destroyed landscapes, monuments in ruin and once thriving now silent liminal wastelands.
Outside of Art specifically, I am working with a few different projects, building a startup, developing Community networks and running a Twitter Spaces show called The New Renaissance (Wednesdays at noon Pacific), defining art culture in a web3 world. All of these things are in the direction of community building and working together with other elite AI artists in the space to become the best, help others become the best, and solidify our position in this brand-new field of artistic exploration and development.
“We were wrong” by Black Label Art Cult
[AI Art Weekly] What does your workflow look like?
My workflow is based off of prompt engineering. As I mentioned before I had developed my own proprietary AI system and with that 99% of my workflow was prompt crafting. I feel like I got a sense of crafting prompts and utilizing that process to get exactly what I’m attempting to achieve. I’ve been jokingly referred to as a “prompt purist”, while others may utilize a process of compositing or multiple runs of the same image. Even though I always have post-process work that is done when it comes to finalizing pieces for mint, the majority of my process is done pre-rendering by crafting exactly what I am expressing with the prompt itself. I typically spend upwards of 10 hours for a final minted piece, taking into consideration all the time spent crafting, curating, and defining to get to that point.
My tip for others is: Create Art. First and foremost. Focus on your expression. Define that, refine that, then express with purpose. AI is just a tool for artistic expression, it is not the artist. Don’t focus on building the perfect prompt, build on developing the perfect expression of who you are on the inside, what you want to tell the world, and who you want to become through your work. Only then will you step from being a prompt engineer, to becoming an Artist.
[AI Art Weekly] What is your favourite prompt when creating art?
This is a tough question because my prompts change constantly as I’m sure they do with everyone else, but in general, my favourite prompt would be one that achieves exactly what I’m intending to express. I believe that a lot of my work resembles a sense of Quasi-realism, I always look for the ability to have the viewer question whether or not what they’re looking at is real, questioning if it’s maybe a photograph of a real scene, while also knowing that it’s not. I typically work with publication references for purity of composition, and I also test atmospheres, and other environmental references. The strongest theme throughout most of my work is a the idea of conceptual contrast. Good versus Evil, Beauty versus Horror, Magnificent versus Ruin. Contrast is an important element for me and what I want to express to the world. I want the viewer to feel both sides of a concept, embrace it, and question it themselves. If that occurs in the mind of those witnessing my work, I believe I have been successful.
“The Cure of Punishment” by Black Label Art Cult
[AI Art Weekly] How do you feel AI (art) will be impacting society?
The evolution of AI and AI Art will impact society as an absolute massive moment of change, and I believe it already has. It will begin to be implemented into every single type of Creative Medium, from art, to 3D, to writing, to the news, to books, to content creation–, to everything. It is creating a scalable process for new creative content to be developed in nearly real-time, at a volume and quality that has never been seen before in the history of the human race – and it’s only just begun.
[AI Art Weekly] Who is your favourite artist?
I have always been partial to Vassily Kandinsky, even as a kid, I always loved his work. Most artists that I admire are abstract expressionists, like Joan Mitchell, and Willem DeKooning. The artists alive today that I most admire is Secundino Hernadez.
Improvisation. Deluge, 1913 by Vassily Kandinsky
[AI Art Weekly] You recently sent an Open Letter about AI art addressing media outlets, can you tell us a bit more about this?
This video was produced out of a need that I saw in the community, after weeks and months of seeing creators being attacked for utilizing AI tools, it became apparent to me that this was happening mostly because the media narrative tends to portray AI in a very negative light, which then leads to people not understanding what AI actually is, which then leads to people fearing it and acting out in a vitriolic manner.
After much discussion with other leaders in the community I decided that it was time we took a stand together, and, at much risk to myself, I decided to take the torch myself and step up with that video for the community as a whole. The amount of support that I’ve received since then has been unbelievable from the AI art community, I have been told that it speaks for them which both humbles and empowers me. When a group of people become voiceless, sometimes it takes a bold act in order to be heard. Since then I’ve been reached out to by a large Web3 publication and had an interview with them, which will be published this week or next. My hope is that it is the first step to heal the community and start a conversation around what I believe to be a better narrative.
[AI Art Weekly] Anything else you would like to share?
The only other thing that I’ll share, is that we’re all in this together. This is a brand new technology, something that is evolving extremely fast, faster than most people outside of this community can possibly comprehend. With that comes a lot of growing pains, and a lot of potential pitfalls, if we don’t stay close and work together as we grow. There’s always a natural state of competition within groups of people, especially artists, and so the one thing that I will continue to preach is that the community is what will bind us all together for success, as we collectively become The New Renaissance.
Each week we share a style that produces some cool results when used in your prompts. This weeks style is retro scifi art style. There is also a dedicated Stable Diffusion model to produce more fine tuned results.
Creation: Tools & Tutorials
These are some of the most interesting resources I’ve come across this week.
DreamArtist looks very similar to MidJourney’s v4 img2img feature. With just one training image DreamArtist learns the content and style in it, generating diverse high-quality images with high controllability.
@Nitrosocke did it again and created the first Multi-Style Model trained from scratch! This is a fine-tuned Stable Diffusion model trained on three artstyles (archer style, arcane style or modern disney style) simultaniously while keeping each style separate from the others.
Two weeks ago I shared a fine tuned Stable Diffusion spritesheet model in this section. Now @ronvoluted created a HuggingFace demo that turns the generated pixel characters into gifs.
Do you want to acquire a deeper understanding of how diffusion models work? Look no further. In this fantastic hands-on Google Colab notebook by @johnowhitaker you’ll learn about diffusion loops, text embeddings, img2img and arbitrary guidance.
This nifty algorithm optimizes your text-to-image prompts by rating the output based on its aesthetics. Notebook by @Omorfiamorphism.
An “older” guide by @KaliYuga_ai where she goes through the process of building an image dataset and turning it into a custom clip-guided diffusion model.
“Haven of the New Renaissance, expressionism, retro scifi art style” by me
And that my fellow dreamers, concludes yet another AI Art weekly issue. Please consider supporting this newsletter by:
- Sharing it 🙏❤️
- Following me on Twitter: @dreamingtulpa
- Leaving a comment on Product Hunt
- Using one of our affiliate links at https://aiartweekly.com/support
Reply to this email if you have any feedback or ideas for this newsletter.
Thanks for reading and talk to you next week!
– dreamingtulpa